Efficient and Numerically Stable Sparse Learning cs高效和数值稳定的稀疏学习政务司司长.uic.pptVIP

Efficient and Numerically Stable Sparse Learning cs高效和数值稳定的稀疏学习政务司司长.uic.ppt

  1. 1、本文档共20页,可阅读全部内容。
  2. 2、有哪些信誉好的足球投注网站(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
  5. 5、该文档为VIP文档,如果想要下载,成为VIP会员后,下载免费。
  6. 6、成为VIP后,下载本文档将扣除1次下载权益。下载后,不支持退款、换文档。如有疑问请联系我们
  7. 7、成为VIP后,您将拥有八大权益,权益包括:VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
  8. 8、VIP文档为合作方或网友上传,每下载1次, 网站将根据用户上传文档的质量评分、类型等,对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档
查看更多
Efficient and Numerically Stable Sparse Learning Sihong Xie1, Wei Fan2, Olivier Verscheure2, and Jiangtao Ren3 1University of Illinois at Chicago, USA 2 IBM T.J. Watson Research Center, New York, USA 3 Sun Yat-Sen University, Guangzhou, China Applications Signal processing (compressive sensing, MRI, coding, etc.) Computational Biology (DNA array sensing, gene expression pattern annotation ) Geophysical Data Analysis Machine learning Algorithms Greedy selection Via L-0 regularization Boosting, forward feature selection not for large scale problem Convex optimization Via L-1 regularization (e.g. Lasso) IPM (interior point method) medium size problem Homotopy method full regularization path computation Gradient descent Online algorithm (Stochastic Gradient Descent) Rising awareness of Numerical Problems in ML Efficiency SVM, beyond Optimization black box solver Large scale problems, parallelization Eigenvalue problems, randomization Stability Gaussian process calculation, solving large system of linear equations, matrix inversion Convergence of gradient descent, matrix iteration computation For more topics of numerical mathematics in ML, See : ICML Workshop on Numerical Methods in Machine Learning 2009 Stability in Sparse learning Iterative Hard Thresholding (IHT) Solve the following optimization problem Incorporating gradient descent with hard thresholding Stability in Sparse learning Iterative Hard Thresholding (IHT) Simple and scalable With RIP assumption, previous methods [BDIHT09, GK09] shows that iterative hard thresholding converges. Without the assumption of the spectral radius of the iteration matrix, such methods may diverge. Stability in Sparse learning Gradient Descent with Matrix Iteration Stability in Sparse learning Mirror Descent Algorithm for Sparse Learning (SMIDAS) Stability in Sparse learning Elements of the Primal Vector is exponentially sensitive to the corresponding elements of the Dual Vector Stability in Sparse learning Example Suppo

您可能关注的文档

文档评论(0)

chenchend + 关注
实名认证
文档贡献者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档